This item by Andrew Levison was originally published on September 8, 2010.
Download pdf Version

In recent days, as increasingly negative projections regarding the November election have appeared, a substantial number of Democrats have been seized with a genuine sense of panic. Many political commentaries have tended to suggest that what is happening may not be just the result of structural factors like the lower participation of pro-Obama groups in off-year elections or the deep recession. Rather, they suggest that a major shift in basic attitudes is occurring – that many Americans are now shifting their allegiance to the Republicans and abandoning Obama and the Democrats. Many Democrats have a sinking fear that support for Obama and the Dems is somehow collapsing.
In order to seriously evaluate this view we have to begin by recognizing that the raw data collected in opinion polls does not come with its own built-in framework for interpretation. Rather, most political opinion poll data is cognitively “shoehorned” into one of two distinct mental models: the “horse race” model or the “sociological” model.
The horse race model is based on the image of two candidates in competition for office and assumes that most voters are continually listening to and evaluating information about the candidates and are therefore very strongly influenced by campaign events like party conventions and televised debates as well as by the daily news headlines. In one formal model in political psychology — called the “online processing” model — voters are visualized as keeping a constantly updated running tally of their impressions of both candidates.
Most national political commentary implicitly accepts the horse race model and generally describes voters as though they were indeed constantly reviewing and revising their impressions and evaluations of candidates and policies. In consequence the ups and downs of candidate approval or voting intentions measured in opinion polls are assumed to be a real-time reflection of this ongoing process.
The sociological model, on the other hand, visualizes a voter’s political attitudes, including decisions about which political party or candidate to vote for, as to a substantial degree determined by an interlocking set of basic value systems that are acquired during childhood socialization and which are then used to determine what the person considers “good” or “bad” and “right” or “wrong.” Once any particular candidate, policy or issue is clearly labeled, categorized and judged within a person’s network of basic value systems, the process of then deciding whether or not to support the candidate or express approval of a particular policy is essentially automatic. Change in these value-based attitudes occurs slowly if at all.
A person’s basic value systems are inherently and inescapably rooted in his or her specific culture and, after the 2000 election, political commentators became very sharply aware of the deep social division of America into the two distinct cultures of “Red vs. Blue” America – the “Red” America that tended to be white, male working class, rural, small town and southern vs. the “Blue” America that tended to be urban, coastal, educated, female and non-white. Numerous commentators noted that these two cultures had very distinct value systems that shaped the evaluation of particular candidates, political parties, polices and issues in dramatically different ways.
In academic political science there is vast literature that studies the demographic and social roots of attitudes like political partisanship, views about issues and candidate choice and few if any political commentators would seriously deny the importance of these underlying social and demographic factors. But, as a practical matter, most daily and weekly political commentary adopts a purely horse race model in which voters are implicitly treated as if they were completely autonomous decision-makers who are reacting entirely to the latest political events.
This approach is understandable since political commentators necessarily try to focus on what is new and novel. The drawback, however is that this perspective can also induce very severe tunnel vision. It needs to be balanced by also looking at current opinion data from a large-scale sociological perspective as well.

For example, from the sociological perspective the decline in Obama’s popularity that occurred between the inauguration and June-July of 2009 can most easily be interpreted as a result of the fact Obama was an unusually unfamiliar figure when he took office – certainly in comparison with someone like Hillary Clinton– and voters were extremely uncertain about how and where to classify him within their basic value systems. Obama’s early appointments and calls for bipartisanship prolonged this uncertainty and many voters did not reach a final conclusion about him until June or July of 2009 when the ambitious scope of the health care bill finally convinced them that Obama was indeed basically a liberal or progressive Democrat.
Once they did reach this conclusion, however, the underlying Red vs. Blue sociological divisions quickly reasserted themselves and Obama’s popularity returned to around the 52.7% he had received in the election. A July 2009 analysis by Dr. Alan Abramowitz showed that, at that time, Obama continued to hold virtually all his support among the demographic groups that had voted for him eight months earlier.
And since July of 2009 Obama’s job approval rating has actually been quite stable.

A gradual downward drift is visible, but it is not a decline that someone who is looking at long term trends would ever interpret as a “collapse.” In fact, what the decline has done is to essentially lower Obama’s job approval rating down to the same level as the percentage of voters who approve of his health care reform legislation and who express general support for Democratic candidates rather than Republicans – a level somewhere in the mid-40’s. The most direct interpretation of this pattern is simply that Obama has retained the support of the voters who firmly share his basic social and political values but has lost the support of those who do not. This may be considered a disappointment from a political marketing and communication point of view but from a sociological perspective it is hardly a surprise.
But why then do so many commentators seem to suggest the opposite recently – that major shifts are occurring in attitudes and that large numbers of Obama’s former supporters are now sharply turning against him?
The main reason is that most political commentary rather mechanically applies a horse race model to the analysis of Obama’s job approval rating – automatically treating any variations in approval and disapproval in the polls as if they directly represented the voters’ running mental tally of their impressions of two candidates in a horse race.
Here’s the chart of Obama’s job approval as you usually see it presented in political commentaries:

Notice how the presence of two converging lines – one for approval and one for disapproval – automatically suggests a horse race paradigm. Most people will almost involuntarily process this chart as showing one horse (in this case the number of people who disapprove of the job Obama is doing) “catching up with” and then surpassing the other.
Note also that this perception is strongly amplified by the narrowing of the scale on the left from 0-100 down to 20-80 — which is what most charts in newspapers and magazines do in order to focus the readers’ attention on the area where changes are occurring. This “focusing in” on a smaller range very substantially increases a readers’ unconscious impression that the changes the chart shows are large, significant and meaningful. (If the chart were shown at 0-100, the two lines would look almost horizontal; at 65-30 (a typical setting) the Democratic line would descend almost all the way from the top to the bottom of the chart.)
Thus, one conclusion is evident. Without a broader sociological perspective to balance it, a horse race perspective will inevitably tend to give a one-sided impression of trends in public opinion. And, moreover, there are also two specific problems with using the horse race perspective to look at job approval ratings in particular.
First, the ups and downs of Obama’s job approval ratings simply don’t directly correlate with the major news events and political debates. During the 75 day period from mid-May to July 2009, when Obama’s job approval fell most precipitously, Nate Silver made a very serious attempt to match any motion in Obama’s approval rating with news or events regarding a whole series of major issues – the economy, the health care debate, and even smaller debates like the Sandra Sotomayer nomination. He found that none of these factors directly corresponded with the week to week changes in Obama’s job approval.
Second, a politician’s job approval ratings are not the same kind of data as a voters’ evaluation of candidates in a two person race. In a two person race, voters are evaluating and comparing two very different individuals with each other. With the job approval rating of a single individual, on the other hand, the percentage of those with an opinion who disapprove of the person is simply the mirror reflection of the approval rate. If 60% of those with an opinion say they approve of the job Obama is doing, then the other 40% of those with opinions necessarily replied that they disapprove. Thus, either one of the two lines on the chart actually contains all the real empirical data. The second line simply presents a redundant negative mirror image of the same information.
(Note: under certain circumstances the Pollster.com charts shown above might not display this relationship exactly because their totals also include those individuals who do not have an opinion. As it happens, however, in the particular charts under discussion here, the “without an opinion” group happens to be extremely small and very stable over the period in question so the two lines do appear to be very close to mirror images of each other, just as they would if the comparison was strictly limited to those with opinions.)
As a result, it is hard to avoid the conclusion that the horse race model is simply an inappropriate way to visualize job approval data. Thinking logically, it just does not make sense to conceptualize a politician’s approval rating as being somehow in a horse race with its own reflection. The first chart is therefore actually a more meaningful representation of the data and what it clearly indicates is that Obama’s job approval is simply not “collapsing” or “crumbling” in any meaningful sense.”
Now keeping the general distinction between the “horse race” and “sociological” perspectives in mind, let’s also look at the generic congressional ballot which asks respondents to choose between unnamed Democratic and Republican candidates. In the last two weeks this data has been the source of tremendous Democratic consternation and dismay.
On the one hand the horse race model is obviously somewhat more applicable in this case since there are indeed two competing groups being compared rather than an evaluation of a single person. But there are still problems.
Here’s how the generic ballot data is usually presented, with the scale on the left once again constricted — this time to a typical 10-65 — and the trend lines for both Democrats and Republicans simultaneously shown.

The chart appears to show the two parties running neck and neck with first the Democrats and then the Republicans pulling into the lead. Once again, the visual presentation powerfully predisposes the reader to think in terms of the horse race paradigm.
But, wait a minute. The idea that people reconsider their basic party affiliation on a week-to-week basis runs counter to virtually everything we know about political partisanship and even at first glance it is clear that the movement in these views is limited to a relatively small group, even though we are now looking at trends going all the way back to election day. And, even more dramatically, when we step back and look at the trends at a full 0-100 scale and also view the trends for Democrats and Republicans separately, the impression that there is any sociologically significant change going on largely disappears.
Here is the chart for the Democrats:

And here are the Republicans:

An extremely slight decline in the Democratic total can indeed be seen in August — if one looks very carefully — and a corresponding rise in the Republican share. Although very small, this change is enough to provide useful data for political scientists making mathematical “generic ballot to congressional seat change” projections. But from a long term perspective, the truth is that the very slight changes in these trend lines over time sure look an awful lot like statistical noise.
Now let’s be clear. The horse race model does indeed have value for many purposes – it is, after all, the main conceptual tool used in all political campaigns. But to evaluate broad social trends like the suggestion that support for Obama and the Democrats is “collapsing,” the horse race perspective must be paired with a broader sociological perspective, certainly before giving way to panic.
In broad, long-term perspective the data simply do not suggest that Obama or the Democrats are in political “free fall.” The fact that the elections this November may shift the balance of power in Congress is basically due to the fact that America is still essentially the same “50-50 nation” that it was in 2000. As a result of this sociological reality extremely small changes – such as a lower participation of pro-Obama demographic groups relative to others in off year election — can easily shift control of Congress.
If Republicans do win control of the House this fall it will be because of the painfully close political balance of power between the two cultures and two political coalitions in America at this point in history and not because major shifts are occurring in the attitudes of the electorate.
To be sure, this will not make the pain of any Republicans victories in November any easier to tolerate, but it does help substantially to keep a balanced view.
So Democrats, step back, take a deep breath and look at the big picture. Obama is attempting to achieve significant progressive reforms with which substantial numbers of Americans do not yet agree. The long-range demographic trends run in our favor, but for the short and medium term, furious, powerful and very substantial opposition is inevitable. The message Democrats should be deriving from the data is the need for firm resolution, renewed commitment and fierce determination – and certainly not justification for demoralization, panic and dismay.